Well, up to this point in time (with some very note worthy exceptions) third party games on the Wii have been rushed ports designed to cash in on the Wii's unexpected popularity. It is likely that at E3 we will see far more actual "Wii games" from third parties and (potentially) by E3 2008 you might see many developers designing with the Wii in mind and then upgrading the games for the PS3/XBox 360.

This is jsut PR talk. Example developers all say the Wii has great motion sensing, but currently we are not thinking of doing any wii projects. This is EA's way of saying we doing games for ps3 because well heres a good reson to justify why we are.

I don't get it, GTA IV is a AAA for adding more graphical powers, mayby bigger cities and so on.

MySims adds the potentional to build everything, your are not any longer just buying stuff. Your are adding an extra layer of complexity so why can't it be seen as something new?

Another example is civilization, which is a AAA title (and your are speaking with a real civ geek here), it has never pushed the boundaries. It have never been the prettiest game nor the most complex, but it is very very good because of that it is enough.

Even if Guitar Hero cost very little to do, why isn't it a AAA title? Since when has price with the quality? Isn't about how well done the game is done? One good example is Little Big Planet, which actual has quit a low budget but most clearly can be one of the reasons to remember this generation with foundness.

And if i understood what kars is saying, he means that since data must go through PPE, the code have to be something that the PPE understands. This you can think being data for single processor. Without optimizing, PPE just runs data to SPE:s without a single SPE being tasked to certain kind of data. Did i go wrong somewhere?

No, the code for the PPE and SPEs are compiled separately.  Kars is saying that the development approach is to program everything on the PPE, test it, work out all the bugs, then start "porting" the important bits of code to the SPEs to improve performance.  It's much easier to find bugs when you're only running code on one core.

Sorry, I already have worked on the Cell. You must not see it as a plattform with 8 processors. The SPUs are no full processors but more like special dedicated coprocessors. While the SPUs can interact with each other, to work as a pipeline for example I don't think that this kind of approach is really advisable in environments, where reaction time is critical, like in games.

SPUs are no typical cores of a multi-core processor, instead they have their own needs and capabilities. I would normally agree with you that for performance reasons you would like independend cores, that interact with each other, although debugging would get VERY hard, but at least in my understanding this was never the world of the Cell. Things like autobalancing that happen in multicores all the time, are not easily available.

The Cell was designed with special applications in mind (low cost cluster computing, one chip multimedia workflow), but never for general purpose applications. If they don't work primarily on their local memory they get quite inefficient. So you are much more fixed for small processing intensive tasks, while a direct interaction with multiple other SPUs becomes difficult, it gets much easier if they use the PPE as a kind of "man in the middle". I aggree, this does not sound very effcient, but I think this is the best approach on a Cell. In my opinion the Cell is not a processor, that you should put into a gaming console.

Sure, I didn't forget it, but in my experience such code has the tendency to have more than two choices (a simple switch case structure is enough), so SMT will only have a limited effect. Till you figure out which branch was the right one you already found the next branch.

 Yes,such problems are very common. It is much easier to talk about the advantages of parallel programs, but when you have to write one, you have to recognize, that these parallel systems can only work efficiently if they are synchronized to each other. This makes them so tricky, together with their debugging problems, which makes the development quite expensive.

So in many cases it is simply cheaper to use only one system instead of a parallel system.

What's Nintendo got beyond this point that indicates third parties are really trying any harder going forward than they have already? At E3, I'll be told to wait til Spaceworld. GTA4's the type of game that takes years to develop and tens of millions of dollars. If it doesn't sell well, the company releasing it will be hurting badly. If MySims doesn't sell, nobody will care.